大模型格式化输出问题（1）

p 评论 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
import os
from langchain_core.output_parsers import JsonOutputParser
from langchain_core.prompts import PromptTemplate
from langchain_core.pydantic_v1 import BaseModel, Field
from langchain_openai import ChatOpenAI

# 设置 OpenAI API 密钥
os.environ["OPENAI_API_KEY"] = "sk-****************"
# 设置 OpenAI API 端点
os.environ["OPENAI_API_BASE"] = "https://apivip.aiproxy.io/v1"

# 初始化模型
model = ChatOpenAI(temperature=0)

# 使用 Pydantic 定义了一个包含 setup 和 punchline 两个字符串字段的数据模型
class Joke(BaseModel):
    setup: str = Field(description="question to set up a joke")        # 描述笑话的开场问题
    punchline: str = Field(description="answer to resolve the joke")   # 描述笑话的结尾答案

    @classmethod
    def model_json_schema(cls):
        # 手动生成 JSON Schema。这是为了兼容 Pydantic v1。
        return {
            "title": "Joke",
            "type": "object",
            "properties": {
                "setup": {
                    "type": "string",
                    "description": "question to set up a joke"
                },
                "punchline": {
                    "type": "string",
                    "description": "answer to resolve the joke"
                }
            },
            "required": ["setup", "punchline"]
        }

# 设置解析器
parser = JsonOutputParser(pydantic_object=Joke)

# 创建提示模板
prompt = PromptTemplate(
    template="Tell me a joke.\n{format_instructions}\n",
    partial_variables={"format_instructions": parser.get_format_instructions()},
)

# 创建链
chain = prompt | model | parser

# 调用链
result = chain.invoke({})
print(result)

极客学堂